Estimate the 3D bounding box from the 2D bounding box,which was detected by the YOLOv4 detector. The training dataset was made by BOXY dataset and KITTI dataset, in order to generate the 3D bounding box in car view.
Rely on:
Pytorch
Argparse
numpy
OpenCV
Usage:
- detect with
bbox_3d_prediction.py
- train with
train.py
- make the dataset with the
kitti2dataset.py
orboxy2dataset.py
Result: